<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-large-xlsr-mecita-coraa-portuguese-clean-grade-3-5
This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0776
- Wer: 0.0632
- Cer: 0.0183
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
19.996 | 0.98 | 31 | 4.5492 | 1.0 | 1.0 |
19.996 | 2.0 | 63 | 3.2872 | 1.0 | 1.0 |
19.996 | 2.98 | 94 | 3.0631 | 1.0 | 1.0 |
6.0801 | 4.0 | 126 | 2.9625 | 1.0 | 1.0 |
6.0801 | 4.98 | 157 | 2.9156 | 1.0 | 1.0 |
6.0801 | 6.0 | 189 | 2.9124 | 1.0 | 1.0 |
3.0002 | 6.98 | 220 | 2.9032 | 1.0 | 1.0 |
3.0002 | 8.0 | 252 | 2.8730 | 1.0 | 1.0 |
3.0002 | 8.98 | 283 | 2.8658 | 1.0 | 1.0 |
2.8967 | 10.0 | 315 | 2.8564 | 1.0 | 1.0 |
2.8967 | 10.98 | 346 | 2.8325 | 1.0 | 1.0 |
2.8967 | 12.0 | 378 | 2.7051 | 1.0 | 1.0 |
2.8241 | 12.98 | 409 | 2.2819 | 1.0 | 0.7857 |
2.8241 | 14.0 | 441 | 1.4363 | 0.9226 | 0.3248 |
2.8241 | 14.98 | 472 | 0.7766 | 0.3399 | 0.0859 |
1.7517 | 16.0 | 504 | 0.5342 | 0.2284 | 0.0565 |
1.7517 | 16.98 | 535 | 0.3976 | 0.2004 | 0.0472 |
1.7517 | 18.0 | 567 | 0.3223 | 0.1593 | 0.0385 |
1.7517 | 18.98 | 598 | 0.2800 | 0.1483 | 0.0359 |
0.7831 | 20.0 | 630 | 0.2448 | 0.1301 | 0.0321 |
0.7831 | 20.98 | 661 | 0.2230 | 0.1225 | 0.0314 |
0.7831 | 22.0 | 693 | 0.2040 | 0.1186 | 0.0306 |
0.5544 | 22.98 | 724 | 0.1883 | 0.1126 | 0.0300 |
0.5544 | 24.0 | 756 | 0.1761 | 0.1131 | 0.0300 |
0.5544 | 24.98 | 787 | 0.1641 | 0.1071 | 0.0281 |
0.3963 | 26.0 | 819 | 0.1593 | 0.1098 | 0.0289 |
0.3963 | 26.98 | 850 | 0.1508 | 0.1032 | 0.0282 |
0.3963 | 28.0 | 882 | 0.1448 | 0.1016 | 0.0275 |
0.357 | 28.98 | 913 | 0.1370 | 0.0934 | 0.0255 |
0.357 | 30.0 | 945 | 0.1333 | 0.0934 | 0.0256 |
0.357 | 30.98 | 976 | 0.1304 | 0.0873 | 0.0244 |
0.32 | 32.0 | 1008 | 0.1232 | 0.0813 | 0.0233 |
0.32 | 32.98 | 1039 | 0.1207 | 0.0835 | 0.0239 |
0.32 | 34.0 | 1071 | 0.1163 | 0.0763 | 0.0229 |
0.2956 | 34.98 | 1102 | 0.1164 | 0.0763 | 0.0217 |
0.2956 | 36.0 | 1134 | 0.1106 | 0.0736 | 0.0209 |
0.2956 | 36.98 | 1165 | 0.1064 | 0.0747 | 0.0205 |
0.2956 | 38.0 | 1197 | 0.1034 | 0.0681 | 0.0194 |
0.2632 | 38.98 | 1228 | 0.1022 | 0.0752 | 0.0202 |
0.2632 | 40.0 | 1260 | 0.1000 | 0.0730 | 0.0200 |
0.2632 | 40.98 | 1291 | 0.1009 | 0.0752 | 0.0204 |
0.2635 | 42.0 | 1323 | 0.0986 | 0.0747 | 0.0210 |
0.2635 | 42.98 | 1354 | 0.0976 | 0.0703 | 0.0203 |
0.2635 | 44.0 | 1386 | 0.0958 | 0.0725 | 0.0194 |
0.218 | 44.98 | 1417 | 0.0930 | 0.0730 | 0.0201 |
0.218 | 46.0 | 1449 | 0.0921 | 0.0681 | 0.0193 |
0.218 | 46.98 | 1480 | 0.0909 | 0.0730 | 0.0201 |
0.221 | 48.0 | 1512 | 0.0928 | 0.0697 | 0.0204 |
0.221 | 48.98 | 1543 | 0.0916 | 0.0681 | 0.0197 |
0.221 | 50.0 | 1575 | 0.0889 | 0.0697 | 0.0196 |
0.2116 | 50.98 | 1606 | 0.0873 | 0.0670 | 0.0192 |
0.2116 | 52.0 | 1638 | 0.0879 | 0.0692 | 0.0197 |
0.2116 | 52.98 | 1669 | 0.0869 | 0.0664 | 0.0193 |
0.2002 | 54.0 | 1701 | 0.0863 | 0.0670 | 0.0197 |
0.2002 | 54.98 | 1732 | 0.0865 | 0.0659 | 0.0190 |
0.2002 | 56.0 | 1764 | 0.0850 | 0.0648 | 0.0184 |
0.2002 | 56.98 | 1795 | 0.0843 | 0.0686 | 0.0187 |
0.175 | 58.0 | 1827 | 0.0844 | 0.0615 | 0.0180 |
0.175 | 58.98 | 1858 | 0.0838 | 0.0637 | 0.0190 |
0.175 | 60.0 | 1890 | 0.0839 | 0.0659 | 0.0189 |
0.1919 | 60.98 | 1921 | 0.0821 | 0.0626 | 0.0183 |
0.1919 | 62.0 | 1953 | 0.0829 | 0.0670 | 0.0186 |
0.1919 | 62.98 | 1984 | 0.0825 | 0.0653 | 0.0191 |
0.1975 | 64.0 | 2016 | 0.0820 | 0.0632 | 0.0191 |
0.1975 | 64.98 | 2047 | 0.0818 | 0.0681 | 0.0193 |
0.1975 | 66.0 | 2079 | 0.0826 | 0.0681 | 0.0192 |
0.1796 | 66.98 | 2110 | 0.0829 | 0.0692 | 0.0194 |
0.1796 | 68.0 | 2142 | 0.0822 | 0.0659 | 0.0193 |
0.1796 | 68.98 | 2173 | 0.0833 | 0.0664 | 0.0194 |
0.1663 | 70.0 | 2205 | 0.0827 | 0.0621 | 0.0194 |
0.1663 | 70.98 | 2236 | 0.0810 | 0.0643 | 0.0195 |
0.1663 | 72.0 | 2268 | 0.0796 | 0.0643 | 0.0189 |
0.1663 | 72.98 | 2299 | 0.0795 | 0.0637 | 0.0186 |
0.1902 | 74.0 | 2331 | 0.0805 | 0.0643 | 0.0185 |
0.1902 | 74.98 | 2362 | 0.0799 | 0.0621 | 0.0183 |
0.1902 | 76.0 | 2394 | 0.0798 | 0.0637 | 0.0179 |
0.1856 | 76.98 | 2425 | 0.0799 | 0.0637 | 0.0179 |
0.1856 | 78.0 | 2457 | 0.0796 | 0.0643 | 0.0184 |
0.1856 | 78.98 | 2488 | 0.0796 | 0.0621 | 0.0184 |
0.1528 | 80.0 | 2520 | 0.0797 | 0.0621 | 0.0185 |
0.1528 | 80.98 | 2551 | 0.0797 | 0.0648 | 0.0191 |
0.1528 | 82.0 | 2583 | 0.0793 | 0.0643 | 0.0186 |
0.1557 | 82.98 | 2614 | 0.0788 | 0.0637 | 0.0182 |
0.1557 | 84.0 | 2646 | 0.0784 | 0.0615 | 0.0181 |
0.1557 | 84.98 | 2677 | 0.0783 | 0.0632 | 0.0176 |
0.1603 | 86.0 | 2709 | 0.0781 | 0.0637 | 0.0185 |
0.1603 | 86.98 | 2740 | 0.0785 | 0.0643 | 0.0186 |
0.1603 | 88.0 | 2772 | 0.0780 | 0.0637 | 0.0183 |
0.1514 | 88.98 | 2803 | 0.0779 | 0.0643 | 0.0182 |
0.1514 | 90.0 | 2835 | 0.0777 | 0.0632 | 0.0181 |
0.1514 | 90.98 | 2866 | 0.0778 | 0.0626 | 0.0184 |
0.1514 | 92.0 | 2898 | 0.0776 | 0.0632 | 0.0183 |
0.1535 | 92.98 | 2929 | 0.0778 | 0.0621 | 0.0180 |
0.1535 | 94.0 | 2961 | 0.0779 | 0.0632 | 0.0181 |
0.1535 | 94.98 | 2992 | 0.0778 | 0.0632 | 0.0183 |
0.1541 | 96.0 | 3024 | 0.0778 | 0.0615 | 0.0181 |
0.1541 | 96.98 | 3055 | 0.0779 | 0.0621 | 0.0180 |
0.1541 | 98.0 | 3087 | 0.0780 | 0.0621 | 0.0180 |
0.1661 | 98.41 | 3100 | 0.0780 | 0.0621 | 0.0181 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3